1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
//! # Not Enough Minerals
//!
//! The solution is [branch and bound](https://en.wikipedia.org/wiki/Branch_and_bound) using
//! a [depth first search](https://en.wikipedia.org/wiki/Depth-first_search) to enumerate every
//! possible combination combined with heuristics to prune those combinations in order to achieve
//! a reasonable running time.
//!
//! The most import heuristic is:
//! * Assume ore is infinite.
//! * Always build a clay robot.
//! * Check if we can do better than the highest score so far in the remaining time, building
//! only geode or obsidian bots.
//!
//! As these simplified rules will always score higher than the real rules, we can immediately
//! prune any branch that can't possibly exceed the current high score.
//!
//! The second helpful heuristic is:
//! * Don't build more bots for a particular mineral than the maximum possible consumption in
//! a single turn.
//!
//! As we can only build one bot per turn, we will never need to generate more resources than
//! that bot can use. For example, if ore robots need 2 ore, clay robots 3 ore, obsidian robots
//! 4 ore and geode robots 5 ore, then the most possible ore robots that we need to build is 5.
//! Any more would go to waste. The same applies for clay and obsidian.
//!
//! The third helpful heuristic is:
//! * Don't build any robot during the last minute
//! * Don't build ore or obsidian robots during the last 3 minutes.
//! * Don't build clay robots during the last 5 minutes.
//!
//! Building any robot during the last minute means that it will be ready *after* the time runs
//! out so it will contribute nothing. The other two rules are corollaries of this rule.
//!
//! For example say we build an obsidian robot with 3 minutes left. It will be ready and collect
//! a resource with two minutes left, which can be spent on a geode robot with 1 minute left,
//! which is too late.
//!
//! Since we only need clay for obsidian robots it doesn't make sense to build clay robots less
//! than two minutes before the cutoff for obsidian robots.
//!
//! The final important optimization is that we don't increment minute by minute. Instead once
//! we decide to buld a robot of a particular type, we "fast forward" in time until there are
//! enough resources to build that robot. This cuts down on a lot of duplicate intermediate states.
use crate::util::parse::*;
use std::ops::{Add, Sub};
/// Each robot generates 1 mineral of a particular type.
const ZERO: Mineral = Mineral::from(0, 0, 0, 0);
const ORE_BOT: Mineral = Mineral::from(1, 0, 0, 0);
const CLAY_BOT: Mineral = Mineral::from(0, 1, 0, 0);
const OBSIDIAN_BOT: Mineral = Mineral::from(0, 0, 1, 0);
const GEODE_BOT: Mineral = Mineral::from(0, 0, 0, 1);
#[derive(Clone, Copy)]
struct Mineral {
ore: u32,
clay: u32,
obsidian: u32,
geode: u32,
}
impl Mineral {
const fn from(ore: u32, clay: u32, obsidian: u32, geode: u32) -> Self {
Mineral { ore, clay, obsidian, geode }
}
/// This is used to compare robot costs so we don't need to check geodes.
fn less_than_equal(self, rhs: Self) -> bool {
self.ore <= rhs.ore && self.clay <= rhs.clay && self.obsidian <= rhs.obsidian
}
}
/// Implement operators so that we can use `+` and `-` notation to add and subtract minerals.
impl Add for Mineral {
type Output = Self;
fn add(self, rhs: Self) -> Self {
Mineral {
ore: self.ore + rhs.ore,
clay: self.clay + rhs.clay,
obsidian: self.obsidian + rhs.obsidian,
geode: self.geode + rhs.geode,
}
}
}
impl Sub for Mineral {
type Output = Self;
fn sub(self, rhs: Self) -> Self {
Mineral {
ore: self.ore - rhs.ore,
clay: self.clay - rhs.clay,
obsidian: self.obsidian - rhs.obsidian,
geode: self.geode - rhs.geode,
}
}
}
pub struct Blueprint {
id: u32,
max_ore: u32,
max_clay: u32,
max_obsidian: u32,
ore_cost: Mineral,
clay_cost: Mineral,
obsidian_cost: Mineral,
geode_cost: Mineral,
}
impl Blueprint {
fn from(chunk: [u32; 7]) -> Self {
let [id, ore1, ore2, ore3, clay, ore4, obsidian] = chunk;
Blueprint {
id,
max_ore: ore1.max(ore2).max(ore3).max(ore4),
max_clay: clay,
max_obsidian: obsidian,
ore_cost: Mineral::from(ore1, 0, 0, 0),
clay_cost: Mineral::from(ore2, 0, 0, 0),
obsidian_cost: Mineral::from(ore3, clay, 0, 0),
geode_cost: Mineral::from(ore4, 0, obsidian, 0),
}
}
}
pub fn parse(input: &str) -> Vec<Blueprint> {
input
.iter_unsigned()
.collect::<Vec<_>>()
.chunks_exact(7)
.map(|slice| Blueprint::from(slice.try_into().unwrap()))
.collect()
}
pub fn part1(input: &[Blueprint]) -> u32 {
input.iter().map(|blueprint| blueprint.id * maximize(blueprint, 24)).sum()
}
pub fn part2(input: &[Blueprint]) -> u32 {
input.iter().take(3).map(|blueprint| maximize(blueprint, 32)).product()
}
fn maximize(blueprint: &Blueprint, time: u32) -> u32 {
let mut result = 0;
dfs(blueprint, &mut result, time, ORE_BOT, ZERO);
result
}
/// Depth first search over every possible combination pruning branches using heuristics.
fn dfs(blueprint: &Blueprint, result: &mut u32, time: u32, bots: Mineral, resources: Mineral) {
// Extrapolate total geodes from the current state in the remaining time.
*result = (*result).max(resources.geode + bots.geode * time);
// Check if this state can improve on the existing high score.
if heuristic(blueprint, *result, time, bots, resources) {
if bots.obsidian > 0 && time > 1 {
next(blueprint, result, time, bots, resources, GEODE_BOT, blueprint.geode_cost);
}
if bots.obsidian < blueprint.max_obsidian && bots.clay > 0 && time > 3 {
next(blueprint, result, time, bots, resources, OBSIDIAN_BOT, blueprint.obsidian_cost);
}
if bots.ore < blueprint.max_ore && time > 3 {
next(blueprint, result, time, bots, resources, ORE_BOT, blueprint.ore_cost);
}
if bots.clay < blueprint.max_clay && time > 5 {
next(blueprint, result, time, bots, resources, CLAY_BOT, blueprint.clay_cost);
}
}
}
/// Simplify the blueprints so that we only attempt to build either geode or obsidian robots,
/// then check that the estimated maximum possible score is greater than the current high score.
/// Additionally we always build a clay robot each turn.
///
/// Since this will always score higher, so we can immediately prune any branch that can't
/// possibly beat the high score.
#[inline]
fn heuristic(
blueprint: &Blueprint,
result: u32,
time: u32,
mut bots: Mineral,
mut resources: Mineral,
) -> bool {
for _ in 0..time {
// Assume ore is infinite.
resources.ore = blueprint.max_ore;
// Only attempt to build geode or obsidian robots.
if blueprint.geode_cost.less_than_equal(resources) {
resources = resources + bots - blueprint.geode_cost;
bots = bots + GEODE_BOT;
} else if blueprint.obsidian_cost.less_than_equal(resources) {
resources = resources + bots - blueprint.obsidian_cost;
bots = bots + OBSIDIAN_BOT;
} else {
resources = resources + bots;
}
// Always build a clay bot.
bots = bots + CLAY_BOT;
}
resources.geode > result
}
/// "Fast forward" in time until we can build a robot of a particular type. This could possibly
/// by the next minute if we already have enough resources.
#[inline]
fn next(
blueprint: &Blueprint,
result: &mut u32,
time: u32,
bots: Mineral,
mut resources: Mineral,
new_bot: Mineral,
cost: Mineral,
) {
for jump in 1..time {
if cost.less_than_equal(resources) {
dfs(blueprint, result, time - jump, bots + new_bot, resources + bots - cost);
break;
}
resources = resources + bots;
}
}